Publications by authors named "Joanna X Liang"
Background: Observational data have suggested that patients with moderate to severe ischemia benefit from revascularization. However, this was not confirmed in a large, randomized trial.
Objectives: Using a contemporary, multicenter registry, the authors evaluated differences in the association between quantitative ischemia, revascularization, and outcomes across important subgroups.
Background: Previous studies evaluated the ability of large language models (LLMs) in medical disciplines; however, few have focused on image analysis, and none specifically on cardiovascular imaging or nuclear cardiology. This study assesses four LLMs-GPT-4, GPT-4 Turbo, GPT-4omni (GPT-4o) (Open AI), and Gemini (Google Inc.)-in responding to questions from the 2023 American Society of Nuclear Cardiology Board Preparation Exam, reflecting the scope of the Certification Board of Nuclear Cardiology (CBNC) examination.
View Article and Find Full Text PDFArticle Synopsis
- - The study focuses on developing an automated system to quantify [18F]-fluorodeoxyglucose (FDG) PET activity in diagnosing cardiac sarcoidosis using deep learning for segmenting cardiac chambers from CT scans.
- - The analysis included 69 patients, revealing that the cardiometabolic activity (CMA) showed the best predictive accuracy for cardiac sarcoidosis, followed by volume of inflammation (VOI) and target to background ratio (TBR).
- - The findings indicate that this automated method provides rapid, objective measurements of cardiac inflammation, showing high sensitivity and specificity for diagnosing cardiac sarcoidosis.
The Registry of Fast Myocardial Perfusion Imaging with Next-Generation SPECT (REFINE SPECT) has been expanded to include more patients and CT attenuation correction imaging. We present the design and initial results from the updated registry. The updated REFINE SPECT is a multicenter, international registry with clinical data and image files.
View Article and Find Full Text PDFArticle Synopsis
- Researchers developed a new AI method to analyze routine CTAC scans from cardiac imaging to create volumetric measurements of various tissues, including fat and muscle, in the chest area.
- The study examined data from nearly 10,000 patients, finding that higher volumes of certain types of body fat (VAT, EAT, IMAT) were linked to an increased risk of all-cause mortality, whereas higher bone and skeletal muscle volumes were associated with lower mortality risk.
- This suggests that CTAC scans hold significant potential for identifying body composition markers that may help predict patient mortality risk beyond their current use.
Background: Previous studies evaluated the ability of large language models (LLMs) in medical disciplines; however, few have focused on image analysis, and none specifically on cardiovascular imaging or nuclear cardiology.
Objectives: This study assesses four LLMs - GPT-4, GPT-4 Turbo, GPT-4omni (GPT-4o) (Open AI), and Gemini (Google Inc.) - in responding to questions from the 2023 American Society of Nuclear Cardiology Board Preparation Exam, reflecting the scope of the Certification Board of Nuclear Cardiology (CBNC) examination.
Article Synopsis
- Transthyretin cardiac amyloidosis (ATTR CA) is gaining attention as a cause of heart failure among older adults, and Tc-pyrophosphate imaging is crucial for diagnosis but is subjective and time-consuming.
- This study tested a deep learning method for automatically measuring Tc-pyrophosphate activity using CT maps, leading to improved efficiency and diagnostic accuracy.
- Results showed that cardiac pyrophosphate activity (CPA) and volume of involvement (VOI) had excellent predictive performance for ATTR CA, correlating with an increased risk of cardiovascular events.
Article Synopsis
- Low-dose computed tomography (CT) scans, used in hybrid myocardial perfusion imaging, provide valuable anatomical and pathological insights beyond just attenuation correction, which may be enhanced through AI-driven frameworks.
- This study analyzed data from over 10,000 patients, segmenting various structures and utilizing deep learning to assess coronary artery health, leading to improved all-cause mortality predictions.
- The comprehensive model integrating data from CT attenuation correction, myocardial perfusion imaging, and clinical factors outperformed other AI models in predicting mortality risk, particularly among patients with normal perfusion.
Article Synopsis
- Chest CT scans are widely used in the U.S., with 15 million performed yearly, primarily for diagnosing various conditions, including cardiac risks.
- A new automated AI system can quickly and accurately assess coronary calcium and various heart chamber volumes from these scans, processing data in about 18 seconds and only missing 0.1% of cases.
- The AI-generated measurements of coronary calcium and heart volumes are effective in predicting overall and cardiovascular mortality, offering a better risk assessment method than traditional evaluations by radiologists.
Article Synopsis
- AI can enhance the analysis of cardiac anatomy from CT-based myocardial imaging, improving the identification of risks related to cardiovascular events.
- A study of over 7,600 patients showed that higher left ventricular mass and volume increased the likelihood of major adverse cardiovascular events (MACEs) by up to 3.31 times.
- Integrating AI-derived cardiac measurements improved risk prediction significantly, as evidenced by a 23.1% better classification in assessing cardiovascular risks.
Article Synopsis
- The study investigates how the size of the heart affects the accuracy of SPECT myocardial perfusion imaging (MPI) in identifying obstructive coronary artery disease (CAD).
- Among 2066 patients, it was found that those with a low left ventricular volume had lower diagnostic performance compared to those with larger volumes, particularly affecting older and male patients.
- The results indicate that smaller heart sizes lead to a significant decrease in the effectiveness of SPECT MPI, highlighting the need for tailored diagnostic approaches based on cardiac size, age, and sex.
Article Synopsis
- Epicardial adipose tissue (EAT) volume and attenuation can indicate cardiovascular risk, but measuring them manually is time-consuming; the study explored using deep learning to automate this process using CT scans.
- Researchers trained a deep learning model on data from 500 patients to accurately identify EAT, achieving results in under 2 seconds compared to 15 minutes for manual analysis.
- After analyzing 8781 patients, results showed that higher EAT measurements were linked to an increased risk of death or myocardial infarction over a median follow-up of 2.7 years, indicating that automated EAT assessments could enhance cardiovascular risk prediction.
Article Synopsis
- Myocardial perfusion imaging (MPI) is widely used to diagnose coronary artery disease, but many patients have normal results; this study explores whether machine learning can identify unique patient profiles among those with normal scans and assess their risk of death or myocardial infarction.
- The research involved a large cohort of over 21,000 patients from an international MPI registry, employing unsupervised clustering to discover four distinct patient phenotypes, revealing differing characteristics and stress testing requirements among them.
- Findings indicated that one specific cluster of patients (Cluster 4), despite having normal scans, faced a significantly higher risk of serious cardiovascular events, suggesting that identifying these phenotypes could enhance risk assessment and patient management.
Article Synopsis
- A new explainable deep learning model has been developed to predict specific cardiovascular risks (like death, acute coronary syndrome, and need for revascularization) based on myocardial perfusion imaging (MPI) combined with clinical data.
- The model was tested with a large group of patients and showed better performance in predicting short-term risks in the first six months post-scan, outperforming traditional methods.
- It provides individualized risk assessments and visual explanations for patients, potentially helping to focus on modifiable risk factors and improve shared decision-making in healthcare.
Article Synopsis
- The study aimed to use unsupervised machine learning to classify patients with known coronary artery disease (CAD) based on their risk profiles during SPECT myocardial perfusion imaging.
- Out of 37,298 patients in the REFINE SPECT registry, 9,221 with CAD were analyzed, identifying three distinct clusters that varied in clinical characteristics, particularly concerning body mass index, diabetes, and hypertension.
- The cluster analysis provided superior risk stratification for all-cause mortality compared to traditional methods based on stress total perfusion deficit, indicating its potential for enhancing patient management in CAD.
Background: Assessment of coronary artery calcium (CAC) by computed tomographic (CT) imaging provides an accurate measure of atherosclerotic burden. CAC is also visible in computed tomographic attenuation correction (CTAC) scans, always acquired with cardiac positron emission tomographic (PET) imaging.
Objectives: The aim of this study was to develop a deep-learning (DL) model capable of fully automated CAC definition from PET CTAC scans.
Direct Risk Assessment From Myocardial Perfusion Imaging Using Explainable Deep Learning.
JACC Cardiovasc Imaging
February 2023
Article Synopsis
- Myocardial perfusion imaging (MPI) is commonly used to assess heart disease risk, but there is a need for better predictive methods, which led to the creation of the HARD MACE-DL model.
- This deep learning model was developed to predict the risk of death or nonfatal myocardial infarction by analyzing various cardiac metrics alongside patient demographics from over 29,000 subjects across several medical centers.
- The HARD MACE-DL model demonstrated superior accuracy in predicting cardiac events compared to traditional methods, achieving a higher prognostic accuracy and excellent calibration in both internal and external validation tests.
Low-dose ungated CT attenuation correction (CTAC) scans are commonly obtained with SPECT/CT myocardial perfusion imaging. Despite the characteristically low image quality of CTAC, deep learning (DL) can potentially quantify coronary artery calcium (CAC) from these scans in an automatic manner. We evaluated CAC quantification derived with a DL model, including correlation with expert annotations and associations with major adverse cardiovascular events (MACE).
View Article and Find Full Text PDFArticle Synopsis
- AI models are effective at diagnosing coronary artery disease (CAD) but can overestimate disease risk due to biased training on high-risk populations.
- A study tested three different training methods, with the third model (using data from low-risk patients) providing the best calibration and predictive accuracy, especially for women.
- Improved AI accuracy in assessing CAD risk is crucial for appropriate patient management, particularly in lower-risk groups where misestimation can lead to unnecessary treatments.
To improve diagnostic accuracy, myocardial perfusion imaging (MPI) SPECT studies can use CT-based attenuation correction (AC). However, CT-based AC is not available for most SPECT systems in clinical use, increases radiation exposure, and is impacted by misregistration. We developed and externally validated a deep-learning model to generate simulated AC images directly from non-AC (NC) SPECT, without the need for CT.
View Article and Find Full Text PDFBackground: We aim to develop an explainable deep learning (DL) network for the prediction of all-cause mortality directly from positron emission tomography myocardial perfusion imaging flow and perfusion polar map data and evaluate it using prospective testing.
Methods: A total of 4735 consecutive patients referred for stress and rest Rb positron emission tomography between 2010 and 2018 were followed up for all-cause mortality for 4.15 (2.
Purpose: We sought to evaluate inter-scan and inter-reader agreement of coronary calcium (CAC) scores obtained from dedicated, ECG-gated CAC scans (standard CAC scan) and ultra-low-dose, ungated computed tomography attenuation correction (CTAC) scans obtained routinely during cardiac PET/CT imaging.
Methods: From 2928 consecutive patients who underwent same-day Rb cardiac PET/CT and gated CAC scan in the same hybrid PET/CT scanning session, we have randomly selected 200 cases with no history of revascularization. Standard CAC scans and ungated CTAC scans were scored by two readers using quantitative clinical software.
Article Synopsis
- The study analyzed the prognostic value of ischemic total perfusion defect (ITPD) in predicting major adverse cardiac events (MACE) in both men and women, using advanced SPECT imaging in an international registry.
- Data from 17,833 patients revealed that ITPD was a significant predictor of MACE, with an interaction indicating differing impacts between sexes; specifically, men had worse survival rates when ITPD was less than 5%, while women had worse survival when ITPD was 5% or more.
- Overall, the findings suggest that moderate to severe ischemia, as measured by ITPD, poses a greater risk for adverse outcomes in women compared to men.
Article Synopsis
- A machine learning model was developed to predict which patients will exhibit abnormal perfusion on myocardial perfusion imaging (MPI) based on clinical information available before tests.
- The model was trained on data from 20,418 patients and tested externally with 9,019 patients, utilizing 30 pre-test features for its predictions.
- Results showed the model outperformed existing clinical models in predicting abnormal perfusion, indicating its potential to improve test selection by physicians.
Article Synopsis
- - Artificial intelligence, specifically deep learning (DL), shows potential to enhance the accuracy of myocardial perfusion imaging (MPI), primarily as a supportive tool for doctors rather than a fully autonomous system.
- - In a study involving 240 patients, physicians’ diagnostic accuracy improved when interpreting MPI with access to explainable DL predictions (AUC 0.779) compared to those who relied solely on standard methods (AUC 0.747).
- - The integration of DL results led to a significant overall improvement in diagnostic performance, with a net reclassification improvement of 17.2%, although the degree of benefit varied among different physicians based on their acceptance of the technology.